← Voltar aos guias

Calling the Claude Messages API: Fixes That Work

What the Claude Messages API Expects

The Claude Messages API uses a stateless request format: your application sends the conversation history with each request, and Claude returns the next assistant message. A working request normally needs a POST request to https://api.59api.com/v1/messages, your 59API key, the Anthropic API version header, and a JSON body containing model, max_tokens, and messages.

At minimum, include Content-Type: application/json, x-api-key: YOUR_59API_KEY, and anthropic-version: 2023-06-01. The messages array should contain objects with a role such as user and content such as a plain text prompt. Use the exact model identifier shown in your 59API account or documentation. Model names can change, and guessing an identifier is a common cause of a 404 or model-not-found response.

A Minimal Request That Is Easy to Test

Before debugging an SDK integration, test the relay with a direct HTTP request. For example, send a POST request to https://api.59api.com/v1/messages with a JSON body similar to this: {"model":"claude-sonnet-4-5","max_tokens":256,"messages":[{"role":"user","content":"Explain HTTP status codes in three sentences."}]}. Add the required headers and replace the model value with one currently available in your account.

A successful response contains an assistant message in the content array. In most integrations, the generated text is read from content[0].text. The response also includes usage information, including input_tokens and output_tokens, which is useful for estimating the cost of repeated prompts.

Common Errors and Practical Fixes

• 401 or 403: Check that the key is active, copied without extra spaces, and sent in x-api-key. Do not put the key in browser-side JavaScript or commit it to a public repository. Store it in an environment variable such as ANTHROPIC_API_KEY.
• 404 or model-not-found: Confirm that the request path is /v1/messages and that the model name is supported by 59API. The Claude family includes Opus, Sonnet, Haiku, and Fable options, but availability and identifiers should be checked in the current service documentation.
• 400 invalid request: Check JSON syntax, make sure max_tokens is a positive integer, and verify that every message has a valid role and content. A conversation should normally alternate between user and assistant messages. System instructions belong in the top-level system field, not as a system role inside messages.
• 429 rate limit: Slow down concurrent requests, add exponential backoff, and retry only transient failures. Keep retries bounded so one incident does not multiply your usage costs.
• Timeouts or truncated answers: Increase the client timeout for long generations, but also inspect max_tokens. A small value can make a response appear incomplete. For streaming, ensure your HTTP client reads the event stream until the completion event instead of parsing only the first chunk.

FAQ: SDKs, Conversations, and Cost

Can I use the Anthropic SDK? Yes, when the SDK allows a configurable base URL. Point it to https://api.59api.com, provide your 59API key, and keep the Messages API headers required by the client. If your library hardcodes Anthropic's endpoint, use its base URL option or send requests with a standard HTTP client.

How do I preserve context? Store the prior user and assistant turns, then send them again in the next messages array. Do not assume the relay stores conversation state. For long chats, summarize older turns to control input-token usage while retaining decisions, constraints, and relevant facts.

Is 59API suitable for production experiments? 59API provides pay-as-you-go access to native official-quality Claude and GPT models through https://api.59api.com, without requiring a large prepaid commitment. That makes it a practical low-cost choice for prototypes, automation, and usage-based services. It is also designed to work with Claude Code, Codex, and OpenAI-compatible SDK workflows, so you can reuse existing tooling where appropriate. Review current pricing, quotas, and operational limits before launching a high-volume service.

When you are ready to test a Claude integration, sign up for 59API, create an API key, verify one direct Messages API request, and only then add retries, streaming, or application-specific tooling. The referral rebate can further reduce the effective cost when eligible.

Preços · Docs